Surface Defect Detection Of Aluminum Casting Based On Machine Vision

In the production process of aluminum die casting, due to various reasons the surface of aluminum casting will appear defects like pores,cracks,scratches and so on.These defects seriously affect the surface quality and mechanical properties of the product,so defect detection during the process of production is very important. At present,the domestic foundry workshop manual use artificial visual to inspection the defects,this method costs high labor intensity and has low work efficiency, the results are vulnerable to human's subjective impact.The machine vision detection technology is non contact, objectivity, high efficiency,high precision, it can effectively overcome the disadvantages of artificial detection method, has become the development direction of the future of industrial detection.In this paper, a set of the aluminum casting defects detection and recognition algorithms is put forward based on the machine vision and image processing technology, the main research work includes:(1) Based on the analysis of all kinds of defects of aluminum casting imaging characteristics, we study on how to extract suspicious regions.Especially for the blowhole defect with small size(diameter ?2 mm), whether if it can be separated will be a difficulty in aluminum casting defect image segmentation research. This paper proposes a segmentation method based on the combination of threshold and morphology,it can accurately segmente defect area,and get the focus of the pixel position.(2)Study on the differencce of real defects(pores, cracks, shrinkage, porosity shrinkage) and pseudo defects(water, skim oil), propose a method to eliminate the pseudo defects base on geometric feature, regional feature and the difference of regional center for grayscale brightness curve.Test shows that this algorithm can effectively reduce the interference of the pseudo defect to the detection results.(3)Study on the expression,extraction and selection of characteristics of aluminum casting surface defect.Through the analysis, the surface defects of aluminum castingscan be expressed by gray, geometry invariant moments and texture features, in order to improve the recognition efficiency and reduce the amount of calculation of the subsequent classification, the principal component analysis is used to reduce the dimension of the feature vector.(4)Research on classifier based on SVM,select the appropriate parameters to establish multi classifier of SVM.Complete the simulation by matlab. Under the condition of ensure accuracy of the case, Reduce the training time and reducethe complexity of vector machine.